BiologicalJournal OJ the Linnean Society (1996), 58: 357-369. With 4 figures

Population genetic variation in rare and endangered IZiarnna () in Virginia

C. NEAL STEWART, JR., GARY ROSSON, BRENDA W. SHIRLEY AND DUNCAN M. PORTER

Department of BioloD, Virginia Pohtechnic Institute and State Univerxip, Blacksburg, VA 24061-0406 USA.

Reckued 28 March 1995, acceptedfor pubhatian 30 August 1995

Random amplified polymorphic DNA (RAPD) markers were used as input for an analysis of molecular variance (AMOVA), homogeneity of molecular variance analysis (HOMOVA), and cluster analysis to describe the population genetic structure of cork, a federally endangered located only in Virginia, and I. rmota, a rare plant in Virginia, Indiana, and Illinois. The analysis was performed to help clarify the taxonomic relationship between the two closely related . We analysed four clones in the only known population of I. cork, breeding stock derived from seeds originating from the population site, and three I. remota populations in Virginia. Eighty-five percent of screened primers revealed DNA polymorphisms in Iliamna. Ninety-nine informative markers were generated using seven primers. No significant statistical differences (at P = 0.05) in RAPD variation was found between species (24% of variance) using the AMOVA procedure. However, within specieslamong populations (31 % of the variance) and within populations (45%of the variance) there were significant differences (Pc 0.002). An unweighted paired group method using arithmetic averages (UPGMA) cluster analysis showed the federally endangered I. cork population to be genetically distinct from the apparently recently introduced (in Virginia: - 100ybp) I. remofa. The lack of significant differences from the AMOVA and the high number shared bands between I. cork and I. remofa suggest that I. cork may be more appropriately classified as a subspecies of I. remofa. in the natural population were genetically similar to one another while the I. cork breeding stock plants and I. remota plants were genetically relatively diverse. 81996 The Linnean Society of London

ADDITIONAL, KEY WORDS: -RAPD - random amplified polymorphic DNA ~ conservation genetics - plant - mallow - AMOVA - analysis of molecular variance - endangered species.

CONTENTS

Introduction ...... 358 Material and methods ...... 360 Sampling strategy ...... 360 RAF'D profiling and statistical analyses ...... 360 Results ...... 36 1 RAPD profiling ...... 36 1 Statistical analyses ...... 364 Discussion ...... 365 Acknowledgements ...... 367 References ...... 368 Correspondence to C. Neal StewartJr., present address: Department of Biology, Univenity of North Carolina, Greensboro, NC 274 12-5001, U.S.A.

357 0024-4066/96/070357+ 13 $18.00/0 01 996 The Linnean Society of London 358 C.N. STEWART ETA.

INTRODUCTION

Zliamna is a North American of the Malvaceae containing seven or eight species. Most of the species are indigenous west of the Mississippi River. However, two species are found exclusively in the east: I. corei and I. rmota. They have showy, insect-pollinated, perfect flowers, and are self-incompatible (T. Wieboldt and J. Randall, unpublished data). Iliumna corei (Shed) SherfT, the Peters Mountain mallow, is a federally endangered plant species that exists naturally in a single population, on Peters Mountain, Giles County, Virginia, USA) (Fig. 1). This population consists of four clumps, each comprising a clone (Stewart & Porter, 1995). The Peters Mountain site is a sandstone cliff (1000m elevation) above the New River. Seeds were unearthed from the site and germinated. The resulting plants were used to establish a breeding population on the campus of Virginia Polytechnic Institute and State University, Blacksburg, Virginia. Iliumna rmota Greene, the Kankakee mallow, is a rare species that was first collected on Altorf Island in the Kankakee River, Illinois (Strausbaugh & Core, 1932). The species apparently has spread eastward along railroad lines to Virginia (Porter & Wieboldt, 1991). In contrast with the mountain- dwelling Z; corei, the Virginia I. remota populations are located near the James River, not unlike the I. remota habitat in Illinois. Since the discovery of I. cork (Strausbaugh & Core, 1932), the proper taxonomic placement of the taxon has been in question. Strausbaugh & Core (1932) placed I.

Figure 1. The sites sampled for mallow DNA in Vuginii, USA. See Table 3 for population abbreviations. RAPD VARIATION IN IWA 359 cord in the same species as I. remota ( remota (Greene) Britton) on the basis of similar morphological features. Shed€ (1 946) revised the classification such that the Kankakee mallow was recognized as I. remota var. ppica and the Peters Mountain mallow was renamed I. remota var. cord Sherff. The basis for this split was a difference in leaf morphology (the Kankakee mallow leaf typically has a broadly triangular terminal lobe subtended by obtuse sinuses, while the Peters Mountain mallow leaf has an oblong terminal lobe with sharp sinuses), leaf size (Kankakee mallow is larger than Peters Mountain mallow), and plant height (Kankakee mallow, l.CLl.7m; Peters Mountain mallow, 0.6-0.9 m). Three years later SherfY(1949)further split the taxa into two species, I. remota Greene and I. cord (She4 SherK, on the additional characters of differing corolla colors (I. cord has a deeper hue), and floral odor (I. remota is scented whereas I. cord is not). Therefore, the division of I. remota and I. cord was based upon very few different morphological characters observed on plants growing in ecologically diverse natural habitats. It is conceivable that the characters used by SherfF in separation of the species are plastic. Plastic traits are defined as those that vary in response to spatial (edaphic and geographic), and temporal heterogeneity, and are known to confound taxonomic classifications (Schlichting, 1986). For example, we have noticed that plants in the breeding population are taller than those on Peters Mountain. Likewise, Bounds (1992) reported that there are statistically significant in situ morphological differences between the population of I. corei and I. remota populations in Virginia. In addition, Strausbaugh & Core (1932) reported the height of I. Cora' on the mountain site approached 2m, although in recent years, the stems have been shorter. Common garden studies would be appropriate to control for differences among Iliamna sites. However, to our knowledge, no data stemming from common garden comparisons between I. cord and I. remota have ever been published, although Sherf€ (1949) and Scott (1973) reported growing them together in such a situation. An alternative to phenotype analysis to delineate taxa is to perform genetic analysis, which is, by nature, not affected by environment. However, there has been limited work using genetic markers in Iliamna. Bounds (1988) reported apparent polymorphisms in five isozymes of I. cord and I. remota with interclonal variation found in I. cord. In a preliminary random amplified polymorphic DNA (RAPD) study, two primers were used to distinguish clones (Stewart & Porter, 1995). No formal population-level analysis has been published to date. Methods have recently been developed to utilize RAPD profiling (Welsh & McClelland, 1990; Williams et al., 1990) in formal population genetic analyses (Lynch & Milligan, 1994; Stewart & Excoffier, 1996). Although numerous DNA polymorphisms may easily by revealed using RAPDs, the markers are primarily dominant, i.e. homozygotes are indistinguishable from heterozygotes (Tinker, Fortin & Mather, 1993; Williams et al., 1993). Stewart & Excoffier (1996) have modified the analysis of molecular variance (AMOVA) technique (Excoffier, Smouse & Quattro, 1992), adapting it for use with dominant markers. In short, they used estimated amounts of autogamy (selfing frequency from 0 to 1.0) to estimate the average genotype frequencies from the phenotypic (RAPD) data. These estimates were then used in the AMOVA. The AMOVA partitions variance to hierarchical levels, e.g. among and within populations, and tests for significance at these predefined levels, similar to an analysis of variance (ANOVA). The objectives of our study were threefold: (1) identlfjr population-level genetic markers in I. cord and I. remota; (2) provide baseline data on the genetic variability of 360 C. N. STEWART ETA. I. corei and I. remota and (3) cia* the taxonomic relationship between I. remotu and I. corei. Our approach was to use RAPD markers in a population genetic analysis. This represents the first use of the AMOVA for RAPDs to analyse an obligately outcrossing plant species, although Stewart & Excoffier (1996) used it for Vuccinium macrocarpon (Ericaceae), a selfing species, and Huff, Peakall & Smouse (1993) used an earlier version of AMOVA, which had not been modified for RAPD data, to analyze obligate outcrossing Buchloe dac&loides (Poaceae). Our approach was to assume that each surveyed population of Iliamna (not including the one I. riuularir sample we included for comparative purposes) were members of a panmictic metapopulation. The AMOVA, a permutational statistics package, tests for population substructuring and will indicate whether significant molecular differences exist within and among local breeding populations. More importantly, the AMOVA will also indicate whether or not significant molecular differences exist between I. remotu and I. corei. The null hypothesis tested is that there is no subdivision at the population and species levels. The associated permutational homogeneity of variance procedure (HOM- OVA) tests whether population genetic heterogeneity is significantly different among populations. This procedure is of particular interest in rare species and conservation biology as it tests whether some populations are genetically more depauperate than others.

MATERIAL AND METHODS

Sampling strategy The I. cord (PM) population (Peters Mountain and Garden) was discussed above, and the clonal structure is presented in more detail in Stewart & Porter (1995). We sampled the three sites in Virginia where I. remotu could be found in June, 1992 (Fig. 1). The largest population was the Mallow Preserve population (MP), located adjacent to a railroad line, and protected by The Nature Conservancy. The population consisted of randomly-spaced clumps of plants, assumed to be single clones. We randomly sampled plants along two transects. Located a few kilometers down-track was the small Iron Gate population (IG), located between a railroad track and road, and across the street from a large factory. The population consisted of only four clumps of plants, all of which were sampled, but only three of which yielded products in the RAPD analysis. This population appeared to be regularly mowed as part of road maintenance. The third population was located alongside Interstate 64.0. It differed from the other two populations in that it was located farther from the railroad line, and consisted of a single 10 m long, 3 m wide oblong clump of plants. We samples every fifth stem along a transect that ran through the length of the clump. A fourth I. remota population was located in Bedford County, Virginia near the James River. However, this population could not be sampled in 1992 because of flooding. For comparative purposes, we also included a DNA sample of I. riuuhrk from Cache County, Utah.

RAPD proJiling and stutiscical anahses

We used fresh leaf tip samples as a source of DNA. DNA extraction, a rapid RAPD VARIATION IN IWA 36 1 miniprep, and RAPD profiling methods are described elsewhere (Stewart & Via, 1993). The raw data for all analyses was a band presence (l)/absence (0) rectangular matrix taken from the composite RAPD profile of each sample (Table 1). Missing data (?) were not analyzed by the AMOVA. The composite profiles were generated from the suite of primers we used in the RAPD reactions: OPA2 (5' TGCCGAGCTG), OPA3 (5' AGTCAGCCAC), OPAl3 (5' CAGCACCCAC), OPA16 (5' AGCCAGCGAA), OPB10 (5' CGTCTGGGAC), CA 947 (5' CCAAC- CACCC), GT 947 (5' GGGTTGGTG). OP primers are from Operon Technologies (Alameda, California), and CA and GT primers were designed by Dr Douglas Rhoades (University of Arkansas-Fayetteville). Genetic distances were estimated using euclidean squared distances as discussed in Huff et al. (1993) and Stewart & Excoffier (1 996) (Table 2). We used a non-parametric AMOVA procedure originally described in Excoffier et al. (1992) as modified by Stewart & Excoffier (1996) to describe population structuring and variability among populations. The associated HOMOVA was used to test for significant molecular variance homogeneity among populations. The I. rivulak sample was excluded from these analyses. In the AMOVA and HOMOVA, we used a hierarchical nested analysis with individuals gathered into populations which were, in turn, gathered into species (Fig. 2). We also performed an unweighted paired group method using arithmetic averages (UPGMA) cluster analysis to produce a dendrogram as a visual aid (Rohlf, 1988). We used the Mantel test to test for goodness-of-fit between a cophenetic (ultrametric) matrix, which was derived from the UPGMA dendrogram and the genetic distance matrix (Mantel, 1967; Rohlf, 1988). This matrix comparison approach is described further in Stewart & Nilsen (1 995).

RESULTS

RAPD projiling

The primary focus of this paper is to delineate the taxonomic of I. cork and I. remota to one another using RAPD markers in a population genetics framework. R4PDs have been used before in the population genetics of rare and endangered species, although analyses have typically been qualitative and not quantitative (see Discussion). The primary attractive factors of RAPDs in conservation studies are the ease of methodologies, abundant polymorphisms, and the small amount of tissue needed for analysis. Indeed, over-collecting by botanists is cited as being an important factor in I. corez's decline on Peters Mountain (Porter & Wieboldt, 1991). In the case of I. corti, only four plants were in existence in nature in 1992, so it was not desirable to sacrifice much tissue. Our methodologies required only a fraction of one leaf per plant. In order to determine whether RAPD profling could be used to characterize genetic variation among Iliamna populations, 42 Werent primers were used to screen four random Iliamna samples. Eighty-five per cent (36) of these revealed polymor- phisms. Seven primers used for the analysis were selected based on the following criteria. The primers had to: (1) reveal polymorphisms, (2) consistently produce strong (brightly staining) amplification products, (3) produce uniform, reproducible markers between replicate PCRs, (4)be insensitive to DNA template concentrations varying from 1 ng/pL to 100 ng/pL (McClelland & Welsh, 1994). Furthermore, we 362 C.N. STEWART ETA. TABLE2. Painvise euclidean squared distances between samples. See Table 1 for abbreviations.

Sample

0 43 0 45 10 0 48990 47 11 13 6 0 39 16 16 17 19 0 43 14 12 13 13 12 0 43 18 16 13 13 16 14 0 44 20 24 17 17 16 14 14 0 44 19 19 18 18 17 15 17 13 0 39 23 23 18 20 22 20 15 22 20 0 39 22 16 17 19 14 14 14 14 13 22 0 44 15 19 16 16 15 13 13 11 10 20 11 0 43 14 18 14 14 15 15 7 11 14 18 11 8 0 43 16 20 15 17 18 16 20 16 13 23 18 15 17 0 43 16 16 15 13 12 6 12 12 13 20 8 7 9 16 0 36 27 23 20 22 21 19 15 17 20 21 15 18 18 21 15 0 38 25 21 22 22 19 21 21 25 26 19 19 22 22 23 17 12 0 43 26 22 25 27 30 26 24 28 27 31 26 27 26 26 26 15 17 0 43 28 24 25 29 26 28 24 26 23 27 24 23 25 24 24 15 15 16 0 45 26 22 23 27 26 28 22 24 25 27 24 23 20 22 24 15 13 16 8 0 9 79 28 26 25 27 30 28 22 26 27 21 28 25 27 26 26 17 11 18 16 14 0 44 27 27 26 28 27 31 25 27 28 28 25 24 22 25 27 19 11 21 19 15 15 0 42 23 21 20 22 25 23 23 29 26 26 25 24 26 25 23 15 11 15 15 15 15 16 0 P 43 31 27 28 30 27 31 27 29 30 32 29 30 28 27 29 22 20 25 13 13 21 18 18 0 43 30 24 23 25 28 26 22 26 29 27 24 25 23 24 24 15 15 18 14 10 16 17 17 16 0 43 33 29 30 30 27 23 27 25 32 29 27 28 26 31 23 18 16 25 23 19 21 22 22 20 11 0 i& 43 26 22 23 27 26 28 22 26 27 23 24 23 22 26 24 15 9 14 14 10 10 13 15 17 12 17 0 il 41 34 30 31 31 28 28 26 30 29 33 24 27 27 32 24 20 18 22 24 24 24 21 19 27 24 23 22 0 i2 40 31 31 26 26 29 29 23 27 28 30 25 26 24 29 25 18 18 19 23 21 21 22 16 28 21 22 19 5 0 i4 41 31 28 29 29 26 26 24 28 27 31 22 25 25 30 22 19 17 20 22 22 22 23 17 27 222120 2 3 0 i5 42 31 27 28 28 27 27 23 29 28 30 21 24 24 29 21 18 16 19 21 21 21 22 16 26 21 22 19 3 4 1 0 i6 42 35 31 30 30 29 29 25 29 30 32 25 28 26 31 25 19 17 21 25 23 23 20 20 28 232221 143 4 0 i7 41 34 30 31 31 28 28 26 30 29 33 24 27 27 32 24 20 18 22 24 24 24 21 19 27 2423220 5 2 3 10 i8 41 34 30 29 29 28 28 24 28 29 31 24 27 25 30 24 19 17 20 24 22 22 21 19 27 2221202323120

Sample riv pm pm pm pm pm pm pm pm pm pm pm pm pm pm pm mp mp mp mp mp mp mp mp mp igl ig2 ig4 il i2 i4 i5 i6 i7 i8 1 2 3 4 A1 A2B1 B2C1 C2aDSD4D6aD6b 1 2 3 4 6 7 8 9 0 w W0, 364 C.N. STEWART ETAL. Variance Sin n ificance

45.4% p<0.002 within populations Figure 2. The nested experimental design for the analysis of molecular variance (AMOVA).The results of partitioning of variance by the AMOVA are shown. Molecular variances withii populations and within species/among populations are significant (P < 0.002). The molecular variance between species is not significant P = 0.27). See Table 3 for population abbreviations.

only scored reproducible fragments (shared fragments between replicate RAPD reactions) that were in the middle molecular weight range (see Penner et al., 1993; Stewart & Porter, in press). Ninety-nine informative markers were generated for analyses. A representative gel showing amplification products using one primer is shown in Figure 3.

Stahtical anabses

The nested AMOVA was used to test the null hypothesis that no genetic subdivision exists among populations or among species. The global analysis showed the molecular variation within and among the four populations tested was significant (asc= 0.43, aST= 0.5 1, P < 0.002), where QSc and aSTare F-statistic analogues (Excoffier et al., 1992). The very high aSTindicates extreme population subdivision (Wright, 1978). This amount of population subdivision is very high compared to the

Figure 3. Typical RAPD gel (0.8% synergel/O.8% agarose composite) stained with ethidium bromide using primer OPA2. See Table 1 for sample abbreviations. RAF'D VARIATION IN ZLL4MNA 365 TABLE3. Painvise aSTbetween populations. pm = I. cm' mountain and garden population mp = I. remotu mallow preserve population; ig = I. motu Iron Gate population; i = I. remot/ Interstate 64 population.

Populations Pm mP ig i

0 0.37 0 0.43 0.08 0 0.62 0.54 0.74 0

mean analogous GST values of other endemics (0.248), long-lived herbaceous perennials (0.2 13), and animal-mediated outcrossers (0.197) (Hamrick & Godt, 1989). We located local differences by performing all possible pairwise comparisons using the same procedure. The aSTvalues between populations showed that the two closest I. remota populations (IG and MP) were not significantly different from one another at the P = 0.05 level (QST = 0.08) (Table 3). This suggests that the smaller IG population possibly originated from the larger MP population at some date following the initial dispersal of I. remota to Virginia. In addition, as, between PM and I. remotu populations (0.47) was about the same as the overall average aST(0.5 1) but was less than that for population I and the other populations (0.63) (Table 3). The HOMOVA tests whether all populations are equally variable (homogeneous variances). The global HOMOVA analysis shows that variance heterogeneities differed among populations (P< 0.002). We located local differences of variance by performing all possible painvise comparisons using the same procedure. Pairwise comparisons at the 0.05 level revealed that the variance of PM (7.12) was not different than MP (7.0) but was significantly greater than that of IG (4.4)and I (1.1). The very low variance of I indicates that this population is nearly monomorphic and is the product of the founding of very few individuals, or is simply one clone varying by somatic mutation. Although there is appreciable genetic variation within both species, it is striking that the variance of the presumably much older PM population is not higher than the young large I. remotu population (MP). Hence, the lack of genetic diversity may explain some of the rarity of I. corei. However, the lack of extensive RAPD variation of I. remota populations may be explained by recent founding events. The UPGMA dendrogram depicts the genomic relatedness of individuals to each other based on RAPD markers (Fig. 4). Discreet populations (PM, I) each form a cluster, and IG and MP individuals intermingle within a single cluster. In addition, the dendrogram provides a very good fit to the triangular euclidean distance matrix (P< 0.001; r = 0.93). Thus, the cluster analysis corroborates the AMOVA by showing three discreet populations (PM, I, IG-MP) from the sample taken. At the species level, I. remota and I. corei each form clusters, but the AMOVA indicates that the genetic differences are not statistically significant.

DISCUSSION

RAPDs have proven to be useful markers in conservation genetics. They have been shown to be roughly equivalent to conventional allozyme markers in measuring genetic diversity (e.g. Liu & Furnier, 1993). That is, when polymorphisms are 366 C. N. STEWART E'TAL. 48.00 36.00 24.00 12.00 0.00

Figure 4. A single UPGMA tree constructed from 99 RAPD characters. Branch lengths indicate relative RAPD similarity (euclidean squared distances are on axis). See Table 1 for sample abbreviations. observed in both systems, the relative genetic variation among individuals and populations is similar. Where low genetic divergence is evident, RAF'Ds reveal polymorphisms when allozymes do not (Brauner, Crawford & Stuessy, 1992; also see Stewart & Excoffier, 1996). Although there have been great strides in developing molecular techniques in population biology, the development of statistical analyses have lagged behind. In most population-level studies, RAF'D data have been treated in a qualitative fashion, and/or been subjected to non-statistical analyses such as cluster analysis (e.g. Castiglione et al., 1993; Hsiao & Rieseberg, 1994; Stewart & Porter, 1995). These approaches are useful in delineating clonal structure and identities within populations, and in determining phylogenetic relationships among taxa. In one example, RAF'Ds were used successfully in clanfjmg the taxonomic relationships among varieties of Ranunculus acniomis (Ranunculaceae) and allied species (Van Buren et al., 1994). Based upon cladistic analysis and morphological variation, Van Buren et al. (1994) elevated R. acrifomis var. aestivalis to R. aestivalis. However, for quantitative population-level analyses in which the investigators want information on population genetic substructuring, they have had to use allozyme markers or other codominant markers. In spite of the dominance problem with RAF'Ds (heterozygotes are indistinguishable from homozygous positives on a gel), some researchers have attempted to use algorithms to describe population structure with RAPDs. For example, Dawson et al. (1993) examined highly inbred Hordeum spontaneum (Poaceae) populations and assumed high marker homozygosity. They used orthodox statistical approaches based on genotype frequency. However, this assumption is generally not appropriate. Russell et al. (1993) used the Shannon diversity index to partition RAF'D variability to within and among population components. Although this provides an RAPD VARIATION IN IMA 367 estimate for population structure, one cannot test for significant differences. In order to use RAPDs in a quantitative fashion one must correct for dominance. Clark & Lanigan (1993) have presented a method of estimating nucleotide divergence with RAPDs. The primary assumption for their analysis is that populations are in Hardy- Weinberg equilibrium. This is often not a valid assumption, especially in small endangered populations. Gibbs, Prior & Weatherhead ( 1994) used this approach in examining populations of snakes. Lynch & Milligan (1994) and Stewart & Excoffier (1996) have recently presented methods to analyse RAPD data in quantitative population-level analyses that require no underlying assumptions of genotype distribution. In the latter treatment, methodologies were presented to allow the estimation of allele frequencies on phenotypic RAPD data using non-parametric statistical procedures. The method presented by Stewart & Excoffier (1996) (AMOVA) was used here to partition the molecular variance into three levels (between species, among populationshithin species, and within populations) based on a priori taxonomic and geographic criteria and tested whether the null hypothesis should be rejected at the three levels. The null hypothesis, that groups of plants are panmictic, can be rejected if molecular characters among groups are significantly different. The null hypothesis was rejected in the among population and within population levels only. However, the null hypothesis was not rejected at the species level (P= 0.27). Therefore, these results offer no evidence in favor of the I. cork and I. remota split (Sherff, 1949) into separate species. Furthermore, the genetic divergence between I. corei and the I. rmota populations is about the same as for all populations, but less than the differentiation between I and the remaining populations (Table 3). So that PM has more similarity at the DNA level to I. remota on average than does population I the other I. remota populations. The UPGMA dendrogram, which has very similar topology to the tree presented in Stewart & Porter (1995),is consistent with the results of the AMOVA with regards to population structure (Fig. 4). However, the AMOVA indicates that the separation of I. cork and I. rmota in the cluster analysis is not significant. These analyses indicate that I. corei should perhaps be recognized as a subspecies of I. remota. But the taxonomic relationship between the two remains enigmatic. Further study encom- passing the other species of Iliamna is needed in order to assess the variation throughout the genus, and to properly classift the endangered I. corei. To this end we will continue to pursue molecular work including sequencing of the ITS rDNA of all species of Iliamna and performing cladistic analyses to further investigate phyloge- netic relationships within the genus. Furthermore, we are seeking to elucidate the molecular basis of I. corez’s ecological demise on Peters Mountain.

ACKNOWLEDGEMENTS

This work was supported by grants from the Virginia Department of Agriculture and Consumer Services to DMP, and a USDA grant to CNS. Thanks to Tom Wieboldt for encouragement, help with locating field sites, and the Iliamna riuularzS accession. We appreciate the help of Suzanne Hill and Mary Lipscomb with sample collecting. We also appreciate the comments on the manuscript by Mary Lipscomb, Tom Wieboldt. Thanks also to Carolyn Hightower for assistance in manuscript 368 C.N. STEWART ETAL. preparation. Access to the Peters Mountain site and Mallow Preserve site near Iron Gate was made possible through The Nature Conservancy.

REFERENCES

Bounds RR. 1988. Patterns of isozyme variation in fliamna cork (She4 SherfE ASB Bulletin 35: 32 (abstract) Boamds RR. 1992. Morphological comparisons of Iliomna populations in Virginia. ASB Bulletin 39: 78 (abstract). Brauner S, Crawford DJ, Steuesy TF. 1992. Ribosomal DNA and RAPD variation in the rare plant family Lactoridaceae. Anmican journal ofBohny 79: 1436-1439. Castiglione S, Wang G, Damiani G, Ban& C, Bisoffi 8,8010 F. 1993. RAPD fingerprints for identification and for taxonomic studies of elite poplar (Populur spp.) clones. Theoretical and Applied Genetics 87: 54-59. Clark AG, Lanip MS. 1993. Prospects of estimating nucleotide divergence with RAPDs. Molecular Biology and Euoluh’on 10: 1096-1 11 1. Dawson IK, Chalmers w,Waugh R, Powell W. 1993. Detection and analysis ofgenetic variation in Hordeurn spontancum populations from Israel using RAPD markers. Molecular Ecology 2: 151-1 59. Excoffier L, Smouse PE, Qwttro LM. 1992. Analysis of molecular variance inferred from metric distances among DNA haplotypes: application to human mitochondrial DNA restriction data. Genetics 131: 474-491. Cibbs HL, Prior KA,Weatherhead PJ. 1994. Genetic analysis of populations of threatened snake species using RAPD markers. Molecular Ecology 3%329-337. HPmrickJL, Godt MJW. 1989. Allozyme diversity in plant species. In: Brown HD, Clegg MT, Kahler AL,Weir BS, eds. Plant Population Genetics, Breeding, and Geneti Rcsources. Sunderland, MA: Sinauer, 43-63. Hsiao J-Y, Rieseberg LH. 1994. Population genetic structure of Tiushania niitakaymnenrir (Bambusoideae,Poaceae) in Taiwan. Molecular Ecology 3: 201-208. HUED& Peakdl R, Smoume PE. 1993. RAPD variation within and among natural populations of outcrossing buffalograss (Buchloe &zc&loides putt.) Engelm.). Theoretical and Applkd Genetics 86: 927-934. Liu 2, Furnier CR. 1993. Comparisons of allozyme, RFLP, and WDmarkers for revealing genetic variation within and between trembliig aspen and bigtooth aspen. “uoretical and Applied Genetics 87: 97-105. Lynch M, Milligan BG. 1994. Analysis of population genetic structure with RAPD markers. Molecular Ecology 3: 91- 99. Mantel NA. 1967. The detection of disease clustering and a generalized regression approach. Cancer Research 27: 209-220. McClelland M, Welsh J. 1994. DNA fingerprinting by arbitrarily primed PCR. PCR Methd and A#luations 4: S5SS65. Penner GA, Bush A, Wise R, Kim W, Doder L, Koshp K, Laroche A, Scoles G, Molnar SJ, Fedak C. 1993. Reproducibility of random amplified polymorphic DNA (RAPD) analysiis among laboratories. PCR Methodr and Apphtionr 2: 341-345. Porter DM, Wieboldt TF. 1991. Vascular Plants. In: Tediger K, Virginia’s Enahgered Species. Blacksburg, VA McDonald and Woodward, 5 1-1 7 1. Rohlf FJ. 1988. .Nunmica1 Taxonq and Mulriumiatc Ana!vis &stem. Setauket, Ny: Eketer Publishing. RussellJR, Hosein F, Johnson E, Waugh R, Powell W. 1993. Genetic differentiation of cocoa (%broma cam L.) populations revealed by RAPD analysis. Molecuh Ecology 2: 89- 97. Schlichting CD. 1986. The evolution of phenotypic plasticity in plants. Annual Rcvinv ofEcology and &stetnutics 17: 667693. Scott MW. 1973. Transplant study of Ilimnna cork and I. remoh in Virginia. Krginiajournal $Science 24: 140. SheaEE. 1946. Notes on certain plants in the Gray’s Manual range. Rhodma 48: 89-97. She& EE. 1949. Miscellaneous notes on dicotyledonous plants. American Journal of Botuny 36: 499-51 1 (abstract). Stewart CN Jr, Excoffier L. 1996. Assessing population genetic structure and variability using RAPD data: Application to Vmcinium macrocarpon (American cranberry).Journal of Euohtionay Biology 9: 153-17 I. Stewart CN Jr, Nilsen ET. 1995. Phenotypic plasticity and genetic variation of Vaccinium macrocarpon (American cranbeny) I. Reaction norms of clones from central and marginal populations in a common garden. Intcmah’onal journal ofpht scimcs 156: 687697. Stewart CN Jr, Porter DM. 1995. The usefulness of RAPD profiling in biological conservation: an application to estimating clonal variation in rare and endangered Zliamna in Virginia. Biological Conrervatton 74: 135-142. Stewart CN Jr, Via LE. 1993. A rapid CTAB DNA isolation technique useful for RAPD fingerprintingand other PCR applications BwTechniques 14: 748-751. Strausbaugh PD, Core EL. 1932. Phymosia remoh. Rhodora 341 142-146. Tinker NA, Fordn MG, Mather DE. 1993. Random ampliied polymorphic DNA and pedigree relationships in spring barley. %retical and Applied GmGtics 85: 976984. RAF'D VARIATION IN IWA 369

Van Buren R, Harper KT, Anderson WR, Stanton DJ, Seyoum S, EnglandJL. 1994. Evaluating the relationship of Autumn buttercup (Ranunculus acrifonnir var. aestivalir) to some close congeners using random ampliied polymorphic DNA. American journal of Bootary 81: 5 14-5 19. Welsh J, McClelland M. 1990. Fingerprinting genomes using PCR with arbitrary primers. Nucleic Acids Research 18: 7213- 7218. Williams JCK, Hanafey MK, Rafalski JA, Tingey SV. 1993. Genetic analysis using random amplified polymorphic DNA markers. Methods in En~mology218: 704-740. Williams JCK, Kubelik AR, Livd w, Rafdski JA, Tingey SV. 1990. DNA polymorphisms amplified by arbitrary primers are useful as genetic markers. Nuclkc AcidF Research 18: 6531- 6535. Wright S. 1978. Evolution and the Gmerics ofPopuhtim. Vol. 4. Variabilib Within andAmongNatura1 Populations. Chicago: University of Chicago Press.